Use default play to predefine column types in update_db()

[mrcaseb]

This resolves #324

Added a new object to sysdata which is a single tibble row that uses random strings for character types with maximum length across the complete dataset (as of 2021).
This should make db drivers to correctly set datatypes in update_db() esp. for long string columns.

It is definitely possible that this needs an update at some point, if one or more of the defined string columns isn't big enough.

Thanks to @friscojosh for the idea. I hope this works.

[tanho63]

Can you prefix this with any::tidyverse? https://github.com/nflverse/nflfastR/blob/master/.github/workflows/pkgdown.yaml#L34

[tanho63]

also just needs to hit redocument to bring in the redact_path arg

[tanho63]

lgtm, maybe josh would be willing to test new methodology

[mrcaseb]

nflreadr sitrep is redocumented. I had to update nflreadr first smh.
I hope the any:: part fixes the tidyverse problem in pkgdown

[tanho63]

encouraging people Seb to always update their packages

[friscojosh]

Testing now!

[mrcaseb]

There is a good chance that non character datatypes are broken now because of the single row principle now. Let's hope it works

That's actually wrong. The first season written to the table is combined with the default play. So we have a complete season for the datatype method to set the types.

[mrcaseb]

Ok, it's definitely broken because I did something stupid, hold on

[friscojosh]

from 4.3.0.9011

── Update nflfastR Play-by-Play Database ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── nflfastR version 4.3.0.9011 ──
• 11:13:53 | Purging the complete data table "nflfastR_pbp_test" in your connected database...
• 11:13:54 | Starting download of 23 seasons between 1999 and 2021...
Error: Data too long for column 'side_of_field' at row 1 [1406]

[mrcaseb]

from 4.3.0.9011

── Update nflfastR Play-by-Play Database ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── nflfastR version 4.3.0.9011 ──
• 11:13:53 | Purging the complete data table "nflfastR_pbp_test" in your connected database...
• 11:13:54 | Starting download of 23 seasons between 1999 and 2021...
Error: Data too long for column 'side_of_field' at row 1 [1406]

That's interesting. So nflfastR calls DBI::dbRemoveTable internally when a complete rebuild is forced. I assume that RMariaDB keeps the field list even when the table is removed.

[friscojosh]

OK, I tried a fresh update to a new table and got the same error.

── Update nflfastR Play-by-Play Database ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── nflfastR version 4.3.0.9011 ──
ℹ 12:17:28 | Can't find the data table "nflfastR_pbp_test2" in your database. Will load the play by play data from scratch.
• 12:17:28 | Starting download of 23 seasons between 1999 and 2021...
Error: Data too long for column 'side_of_field' at row 1 [1406]

However, update_db() works as expected when the table is built manually after binding rows for all available data into one dataframe.